Paper | Published:

Critical evaluation of energy intake using the Goldberg cut-off for energy intake:basal metabolic rate. A practical guide to its calculation, use and limitations

Abstract

OBJECTIVES: To re-state the principles underlying the Goldberg cut-off for identifying under-reporters of energy intake, re-examine the physiological principles and update the values to be substituted into the equation for calculating the cut-off, and to examine its use and limitations.

RESULTS: New values are suggested for each element of the Goldberg equation. The physical activity level (PAL) for comparison with energy intake:basal metabolic rate (EI:BMR) should be selected to reflect the population under study; the PAL value of 1.55 x BMR is not necessarily the value of choice. The suggested value for average within-subject variation in energy intake is 23% (unchanged), but other sources of variation are increased in the light of new data. For within-subject variation in measured and estimated BMR, 4% and 8.5% respectively are suggested (previously 2.5% and 8%), and for total between-subject variation in PAL, the suggested value is 15% (previously 12.5%). The effect of these changes is to widen the confidence limits and reduce the sensitivity of the cut-off.

CONCLUSIONS: The Goldberg cut-off can be used to evaluate the mean population bias in reported energy intake, but information on the activity or lifestyle of the population is needed to choose a suitable PAL energy requirement for comparison. Sensitivity for identifying under-reporters at the individual level is limited. In epidemiological studies information on home, leisure and occupational activity is essential in order to assign subjects to low, medium or high PAL levels before calculating the cut-offs. In small studies, it is desirable to measure energy expenditure, or to calculate individual energy requirements, and to compare energy intake directly with energy expenditure.

Introduction

Over the decades from the thirties to the seventies, the ‘validity’ of dietary assessment techniques has been tested by comparing one method with another, but without any means of testing which, if any, method was valid. In the eighties and nineties, the advent of biomarkers of intake, including urinary nitrogen excretion and doubly labelled water measures of energy expenditure, uncovered the frequent bias to the under-estimation of food intake. It is now generally acknowledged that mis-reporting of intake is a serious problem in studies of nutrition and health.

The phenomenon under investigation is the marked bias to the under-estimation of food intake when the information is derived from self-reported dietary assessments. The word that has been, and still is, widely used for this phenomenon is ‘under-reporting’. This unfortunately carries connotations of causality, suggesting it is due only to conscious or unconscious omission of food items from the report. However, under-eating (dieting) is also a significant contributor to the systematic bias1,2,3,4,5 and over-reporting/over-eating has also been demonstrated.6,7 The phenomenon is therefore more accurately described as ‘providing diet reports of poor validity’. This is a cumbersome phrase and therefore the word ‘under-reporting’ is still used in this paper, but should not be taken to imply causality.

Poor validity of dietary assessments has implications for the interpretation of studies of diet and health. Under-estimation of nutrient intake will inflate estimates of the incidence of inadequate intake.8 Between-subject differences in reporting, whether quantitative or qualitative, will create variable bias that may distort the relationships between nutrient intake and health. That conclusions can be altered by the inclusion or exclusion of ‘low energy reporters' has been demonstrated.9,10

A key problem is how to identify those individuals who have provided data of poor validity. Until that can be done, the characteristics of the bias cannot be fully explored and it is not possible to determine how to minimize the effects in the analysis of dietary data. A number of authors have used the Goldberg cut-off11 to identify ‘low energy reporters' and to explore their characteristics or to examine the effects of low energy reporting on the data and the conclusions to be drawn from it. However, the technique has not always been fully understood or correctly applied. The papers by Goldberg, Black and colleagues were not written as a practical guide to a tool for investigating under-reporting. Thus the information needed to apply the concept was not easy to extract, and the key statistics for the derivation of the equation were relegated to the Appendix. Misinterpretations of the concepts have included using a cut-off calculated for mean intake to identify individual under-reporters, confusing the cut-off for ‘habitual’ intake with that for low intake obtained by chance, and interpreting the illustrative tables of cut-off values based on a physical activity level (PAL) value of 1.55 as recommendations for universal application.

The present paper re-states the principles underlying the Goldberg cut-off, re-examines each parameter in the equation and the values to be inserted, and provides guidance for its application and comments on its usefulness and its limitations.

The principles of the Goldberg cut-off

The validation of reported energy intake (EIrep) rests on the fundamental equation

 Energy intake (EI) = energy expenditure (EE)±changes in body stores

and the assumption that at the group level changes in body stores can be ignored, and therefore

 EI=EE (1)

Good agreement between EI and EE when weight is stable has been demonstrated in an experimental situation.12 Thirty-two healthy subjects were fed on a liquid formula until weight had been stable for 14 days. At stable weight EI was 2536±94 kcal/day and doubly labelled water (DLW) EE was 2564±83 kcal/day. Agreement in a natural situation was provided by two studies in which EI was observed by researchers and EE measured by DLW. The mean difference between EI and EE in 14 hospitalized elderly women was −1% and in 10 men eating ad libitum in a metabolic facility it was +5%.13,14,15

The absolute energy requirements of individuals vary with age, sex and body size, but, since the basal metabolic rate (BMR) also varies with age, sex and body size, energy requirements may be stated in general terms as multiples of BMR.16 This expression of energy requirement (EE:BMR) has also come to be known as the physical activity level (PAL). Thus Eq. (1) may rewritten as

 EI:BMR=PAL

and the mean reported energy intake (EIrep) in any study may be expressed as EIrep:BMR and compared with the expected PAL for that population. However, absolute agreement cannot be expected since there are errors in the measurement of all elements of this equation. The mean EIrep:BMR could derive from valid diet assessments but still be less than (or greater than) PAL. The confidence limits of the agreement between EIrep:BMR and PAL must be determined. The equation for calculating the confidence limits (cut-off) is given below. The full statistical derivation was given in the original paper.11 In that paper, since the energy expenditures in the studies reviewed were unknown, only the lower confidence limit below which the reported mean energy intake was unlikely to represent valid data was calculated. It was calculated assuming a sedentary lifestyle. However, if the activity of a population is known such that it is possible to assign an appropriate mean PAL, then the upper confidence limit may also be calculated.

To determine whether a given mean value of EIrep:BMR in n subjects is acceptable, the following must be satisfied:

where PAL is the mean PAL for the population under study, s.d.min is −2 for the 95% or −3 for the 99% lower confidence limit, s.d.max is +2 for the 95% or +3 for the 99% upper confidence limit, and n is the number of subjects in the study. S is the factor that takes account of the variation in intake, BMR and energy requirements, and is given by

where CVwEI is the within-subject coefficient of variation in energy intake, d is the number of days of diet assessment, CVwB is the coefficient of variation of repeated BMR measurements or the precision of estimated compared with measured BMR, and CVtP is the total variation in PAL. CVtP is the coefficient of variation derived from the mean and standard deviation of a study and includes true between-subject variation, an element of within-subject variation and methodological errors. In the original paper11 this term was designated CVb (between-subject variation), but it has been changed to CVt in the present paper to conform to the nomenclature used in a subsequent paper examining within- and between subject variation in repeat DLW measurements.17

This equation calculates the confidence limits (cut-offs) that determine whether the mean reported energy intake is plausible as a valid measure of food intake even if chance has produced a dataset with a high proportion of days of genuinely low (or high) intake. This was termed CUT-OFF 2 in the original paper. In order to maximize the sensitivity and specificity of the cut-off,7 each element of the equation should be substituted with values appropriate for the specific study being evaluated.

Factors for substitution into the Goldberg equation

Energy expenditure of the population under study: the PAL

In the original paper that examined the reported energy intake from 37 different studies,18 EIrep:BMR was compared with a PAL value of 1.55, the WHO value for ‘light’ activity.16 The reasons for choosing this value were several. First, the energy expenditure or physical activity in the studies examined was unknown; second, choosing too high a value might have exaggerated the extent of under-reporting; third, the WHO recommendations for energy requirements16 were well known and widely used; and fourth, calorimetry and early DLW work confirmed the value of 1.55×BMR for light activity as a probable minimum energy requirement for a normally active but sedentary population (not sick, disabled or frail elderly).11

A subsequent review of 74 DLW studies comprising 574 free-living subjects19 indicated that 1.55 was a conservative value and that many groups have higher energy expenditures. That review suggested the following broad categories:

‘The relationship between lifestyle, activity and PAL suggested by the data can be summarised as follows:

  • Chair-bound or bed-bound 1.2

  • Seated work with no option of moving around

  • and little or no strenuous leisure activity 1.4–1.5

  • Seated work with discretion and requirement to

  • move around but little or no strenuous leisure

  • activity 1.6–1.7

  • Standing work (eg housewife, shop assistant) 1.8–1.9

  • Significant amounts of sport or strenuous leisure

  • activity (30–60 min 4–5 times per week) +0.3

  • Strenuous work or highly active leisure 2.0–2.4n

The data provide little evidence to quantify the energy cost of manual occupations, but the range 2.0–2.4 is suggested as the maximum for a sustainable lifestyle.’

Other PAL values for broad categories are given by the FAO/WHO/UNU Expert Consultation on Energy and Protein Requirement16 and UK report on Dietary Reference Values.20 The WHO report also includes tables showing the factorial derivation of the PAL values. The DLW data from 574 free-living subjects19 were also used to derive mean energy expenditures by age and sex. These are shown in Table 1. However, it should be noted that these subjects were predominantly white collar in origin and very few were identified as having manual occupations. The higher PAL for the age group 18–29 probably reflected more active leisure pursuits, and the lower values in the older individuals, less active leisure. Values for specific groups in the population could be obtained from individual DLW studies. An appendix to the above paper19 listed all 74 studies with details of the subjects and the measured energy expenditures expressed both as mega joules and as PAL. However, such figures should be used with caution since most DLW studies have been conducted on small numbers of selected subjects.

Table 1 Energy expenditure in different age and sex groups

There exists therefore a body of data to suggest suitable PAL values for various population groups. However, knowledge of the activities of the population is required, suggesting that a questionnaire on home, leisure and occupational activity should be included routinely in all dietary surveys.

Between-subject variation in physical activity (CVtP)

CVtP is the coefficient of variation of total variation in energy requirement expenditure calculated from the mean and standard deviation of a study. This figure, although referred to as ‘between-subject’ variation and notated as CVb in the original paper,11 includes within- (CVw) and between- (CVb) subject variation and methodological errors. This is the value to be used when the group mean EIrep:BMR is being evaluated. It is also the value to use when individual EIrep:BMR are being evaluated, since the EIrep:BMR is being evaluated against a population mean that includes both between and within-subject variation. The value of 12.5% used in the original paper was taken from the FAO/WHO/UNU report on energy and protein requirements16 (Section 4.5, p 44), but only one study21 was cited to support this value. The text stated ‘…the measurement of total energy expenditure over a week indicates that the inter-individual variability of expenditure, in a specified group, has a coefficient of variation (CV) of about±12.5% on a body weight basis.21

Accumulated data from doubly labelled water studies (Figure 1) showed a wide range of PAL, from about 1.2 to over 2.0, at all ages.19 The CVtP from these data are shown in Table 1. They ranged from 9.5 to 23.8% in different age–sex groups with a pooled mean of 15.4%. Examination of studies with repeat DLW measurements found the values to be little reduced when CVtP was calculated using subject means derived from more than one DLW measurement.17 A round figure of 15% is suggested as a suitable average value to substitute into the Goldberg equation, but other values may be used if deemed more appropriate for any given study.

Figure 1
figure1

Distribution of individual PAL values from 574 free-living individuals in whom energy expenditure was measured by doubly labelled water. (From Black et al, 1996.19)

Within-subject daily variation in energy intake (CVwEl)

The within-subject day to day variation in food intake (CVwEI) of individuals is large. Figure 2 shows the energy intake of one individual who maintained a weighed diet record every sixth day for 1 y.22 The zero line indicates the mean intake averaged over the year, ie the ‘habitual' intake. Average intakes from 1, 3 or 7 successive days of recording are plotted as the differences from zero. The fine dotted line shows the daily differences and demonstrates the enormous variability from day to day. The 3 day and 7 day average differences show the values obtained from alternative short-term measurements. In the year-long study of dietitians from which these data were taken,23 the within-subject variation in daily energy intake ranged from 10 to 50% in individuals, with a pooled mean of 26%.24

Figure 2
figure2

Daily energy expenditure of one individual undertaking a weighed diet record every sixth day for 1 y, illustrating the variation obtained from 1, 3 and 7 d averages. Mean intakes for 1, 3 and 7 successive days of recording are shown as the difference between that intake and the year-long average (habitual) intake. (From Black, 1999.22)

Table 2 show the CVw from other studies reviewed by Bingham25 and Nelson and colleagues.26 These ranged from 14 to 45% with a pooled mean of 23%. Thus 23% is suggested as a suitable average value to substitute into the Goldberg equation, but a specific CVwEI may be calculated for any given study according to the formula

Table 2 Pooled within-subject variation (CVw) in energy intake

where CVi is the CV calculated for each subject from the number of days of dietary assessment available for that subject, and n is the number of subjects.

Variation in basal metabolic rate (CVwB)

Since energy requirement and intake are expressed as multiples of the BMR, between-subject variation is taken into account and only within-subject variation need be considered in calculating the confidence limits. This includes both measurement error and variation with time on repeated BMR measurements. The classical BMR measurement is made at least 12 h after the last meal, after an overnight sleep in the place of measurement, completely at rest and at thermoneutral temperature (about 23°C). However, practicalities often dictate that these conditions are not fully met, in which case the measurement is called resting metabolic rate (RMR). Either the subject has to walk from the place of sleep to the place of measurement (in-patient RMR) or sleeps at home and travels fasted to the place of measurement in the morning (out-patient RMR), in both of which situations the measurement is done after a rest period of usually 30–60 min. For the out-patient RMR, researchers do not have control over the size of the last meal, length of fast or activity in the previous 24 h all of which influence the RMR. Other factors which may influence RMR and which differ between studies are the nature of the equipment used (face mask, nose clip, ventilated hood, ventilated tent, or whole body calorimeter), subject familiarity with the equipment and procedures, amount of activity on the morning of the measurement, or length of rest period before the measurement, all of which may increase the within-subject variation over that obtained under the well-standardized conditions of the classical BMR.

Under well-standardized conditions CVwB for measured BMR averages about 2.5%. This value was derived from 279 subjects measured in a whole body calorimeter on successive nights on which the antecedent diet and activity had been the same.27 However, in community studies subjects are sampled from the population and variation includes that due to antecedent diet and activity, natural weight fluctuations, menstrual cycle and methodological error. Table 3 shows results from studies that have specifically investigated within-subject variation in BMR; the majority were conducted in young men. The mean CVwB from these 11 studies was 3.9%. Table 4 shows the within-subject variation in BMR obtained from DLW studies with repeat measurements. The mean value from all 12 studies was 5.5%, but the mean from free-living individuals without imposed study conditions was 4.7% (first six studies in Table 4). The mean of these studies and those in Table 3 was 4.2%. A round figure of 4% is suggested as a suitable average value to substitute into the Goldberg equation, but other values may be used if deemed more appropriate for any given study.

Table 3 Within subject variation in measured BMR
Table 4 Within-subject variation in measured BMR from DLW studies

Where no measured BMR or RMR is available, it may be calculated from weight and height or weight alone. Many equations are available and each gives a slightly different result.28 The equations based on the largest body of data and most widely used in dietary assessment are those of Schofield,29 modified to provide equations for age range 60–74 and >74 y.20 The value for CVwB used in the original paper11 was a single value of 8%. However, specific values for the different age–sex groups of the Schofield equations are available29 (Tables Al.l and A1.2, pp 31–32 of Schofield's paper). The standard errors for predictions of BMR expressed as a percentage of the mean BMR (ie coefficients of variation for predictions) from these tables are summarized in Table 5. A figure of 8.5% is suggested as a suitable average value to substitute into the Goldberg equation, but other values may be used for studies of specific age–sex groups.

Table 5 Standard errors for predictions of BMR expressed as a percentage of the mean BMR (ie coefficients of variation for predictions)

The validity of the Schofield equations for obese subjects

The body of data on which the Schofield equations were based contained few subjects of very high body weight (>83 kg for men or >76 kg for women), and thus probably few of very high body mass index (BMI). Since adipose tissue has a lower metabolic rate than lean tissue, the BMR of obese persons may be over-estimated by the equations and thus the extent of under-reporting as measured by EI:BMRestimated associated with obesity also over-estimated. This question has been examined using the database compiled for the review of DLW energy expenditure.19 Figure 3 shows the plot of (BMRestimated−BMRmeasured) against ln(BMI) for males and females aged 18 y and over. The data have been fitted with a quadratic equation. For males there was no significant difference between BMRest and BMRmeas associated with BMI (P=0.4). For females the difference was highly significant (P=0.0001), but only for women with BMI>35 kg/m2 was it of practical importance. Table 6 shows the mean difference by BMI category for both sexes. Since the proportion of the population with BMI>35 kg/m2 is approximately 5–7% only (Health Survey for England 1997, Department of Health press release), it is unlikely that using estimated BMR has caused the association between under-reporting and high BMI to have been exaggerated. The association appears extremely robust; it has been found in many studies including, for example, the USA CSFII 85–86 survey3 and the UK NDNS survey.31 Limited data on the large under-reporting in the massively obese suggest that even adjusting the BMR would not transfer these persons into the category of acceptable reporters. In 324 women with measurements of EI and DLW EE, 18 had BM≥35 kg/m2. In these women mean EI:EE was 0.74 (s.d. 0.33), compared with 0.78 (s.d. 0.25) for those with BMI 25–34 kg/m2 and 0.87 (s.d. 0.22) in those with BMI<25 kg/m2.

Figure 3
figure3

Difference between estimated and measured BMR plotted against 1n(BMI) and fitted with a quadratic equation.

Table 6 Mean difference BMRestimated−BMRmeasured (MJ) according to category of BMI as measured in DLW studies

Using the revised Goldberg cut-offs to evaluate reported energy intake

There are several scenarios for the evaluation of reported energy intake (EIrep).

1. Evaluating the group mean EIrep.

(A) In a small survey where n<100. (B) In a large survey where n>100.

2 Evaluating EIrep at the individual level.

(A) Where intake has been measured by short records or recalls of <14 days.

(B) Where intake has been measured by methods presumed to measure ‘habitual’ intake, ie records of 14 days or more, diet history (DH) or food frequency questionnaire (FFQ).

Each of these situations will be considered in turn.

Evaluating group mean EIrep:BMR in small studies (n<100)

Table 7 illustrates how the CL (cut-offs) for group mean intake vary depending on the factors substituted in the equation. Standard factors for CVtP and CVwEI have been used and the CVwB for estimated BMR. The effects of varying the PAL value chosen as the yardstick, the number of subjects in the study and the number of days of dietary assessment are shown. A number of key points are:

Table 7 The effect of changing the values substituted into the Goldberg equation on the calculated confidence limits (cut-offs): evaluating group mean EI:BMR
  • The choice of the PAL value used as the yardstick for comparison with EI:BMR is important. If 1.55 was used for a population with a true mean energy expenditure of 1.80, then bias might be undetected or the degree of bias under-estimated.

  • Bias is detected more readily in large studies, since the confidence limits (CL) become markedly narrower as n increases, the 95% CL range decreased from 0.38 with n=10 to 0.04 with n=1000.

  • Increasing the days of dietary assessment (d) from 1 to 4 substantially improved chances of detecting bias. The 95% CL range decreased from 0.57 with d=l to 0.41 with d=4. However, the CL for mean EIrep:BMR were not substantially improved by increasing d above 4 days.

Evaluating group mean EIrep:BMR in large studies (n>100)

When evaluating mean intake of larger groups there are further considerations which impact on the study design:

  • Table 7 shows that changing n had a small effect on the confidence limits above n=l00. For larger studies the CL were barely different from the PAL chosen for comparison. Studies with n>100 therefore constitute a special case in which the mean Elrep:BMR may be compared directly with the PAL.

  • Table 7 also shows that changing d, the number of days of dietary assessment, was of little importance compared with changing n, the number of subjects, in determining ability to detect bias in the mean intake. With n>100, increasing d had only a small effect on the CL and a negligible effect when n>500. Thus the number of days of assessment (d) is irrelevant to the evaluation of group mean intake in very large studies (n>500).

Evaluating EIrep:BMR at the individual level in studies with short records or recalls (d<14)

The sensitivity and specificity of the Goldberg cut-off for identifying invalid reports at the individual level was investigated in depth in a previous paper.7 That analysis was based on subjects from DLW studies. The sensitivity of the Goldberg cut-off was assessed as the proportion of subjects classified as under-reporters (UR) by EI:EE who were correctly classified as UR by EI:BMR. Some conclusions from that analysis are summarized below, but the original paper should be consulted for full understanding of the issues.

The evaluation of individual intake is a special case of the Goldberg cut-off in which n=1. Table 8 shows how, in this special case, the CL alter according to changing the variables in the equation.

Table 8 The effect of changing the values substituted into the Goldberg equation on the calculated confidence limits (cut-offs): evaluating EI:BMR at the individual level
  • Confidence limits were substantially wider when n=1 compared with larger n. The 95% CL ranges were 1.09–1.88 in Table 8 compared with 0.04–0.57 in Table 7. Thus the ability of the Goldberg cut-off to detect bias (invalid reports) at the individual level is limited.

  • The ability to detect invalid reports improved as the number of days of dietary assessment (d) increased up to 7 days; the 95% CL range dropped from 1.88 for d=1 to 1.23 for d=7. The improvement beyond 7 days was smaller. (This should not be taken to indicate that 7 days is necessarily a desirable maximum, since the precision of the estimate of individual intake improves with √d and more days might be desirable in a given study)

  • If 1.55 is a conservative estimate of normal activity, then using a higher PAL for comparison might improve the identification of under-reporters. However, using higher PAL values raised both the lower and the upper CL. In the previous paper7 it was shown that if a single PAL and cut-off were used to evaluate the intake of all subjects in a study, about 20% were misclassified, and that the proportion of subjects misclassified remained largely unaltered whether the PAL chosen was 1.55 or higher. As the PAL value was raised, more under-reporters were identified but more ‘acceptable’ reporters were misclassified as under-reporters. The sensitivity was only improved if individuals were assigned to low, medium or high levels of activity7 and three different cut-offs calculated for the PAL values assigned to each level.

  • The ability to detect invalid reports may be marginally improved if a measured BMR is available. In small studies where the integrity of individual data is more important, measuring BMR may avoid misclassifications32 which could distort analysis.

Evaluating EIrep:BMR at the individual level for long records, diet history or food frequency questionnaire

Evaluation of EIrep:BMR when ‘habitual’ intake is assumed to have been measured is another special case of the Goldberg equation. If the number of days (d) is taken as infinity, the expression for CVwEI disappears, but the errors due to CVwB and CVtP remain. The CL for ‘habitual’ intake therefore are virtually the same as those for records of 21 days.

In the original paper,11 it was suggested that ‘habitual’ intake be compared with an EI:BMR of 1.35, the lowest value compatible with a normally active lifestyle. This was referred to as CUT-OFF 1 and reported intakes for ‘habitual’ intake below this value were to be rejected. However, this strategy ignored the errors associated with CVwB and CVtP and is inconsistent with the use of the Goldberg cut-off as described above. It has also contributed to confusion in the use of the Goldberg cut-off. CUT-OFF 1 should be abandoned.

Limitations of the Goldberg cut-off

There are considerable limitations to the Goldberg cut-off. It has poor sensitivity for defining invalid reports at the individual level. The CL are wide and only extreme degrees of mis-reporting can be identified. It can make no distinction between varying degrees of mis-reporting.

The major limitation is that the use of EI:BMR for evaluating energy intake depends on knowledge of energy requirements or energy expenditure. For good reasons the original paper11 assumed a sedentary lifestyle and used a PAL of 1.55 as the yardstick for evaluating mean energy intake. Since that publication numerous authors have used the cut-off for 1.55 and n=l to identify individual low energy reporters (LERs). This strategy has enabled the characteristics of LERs to be explored and enhanced understanding of bias in reporting. However, analysis of accumulated data from DLW studies has shown, first, that 1.55 is a conservative value for normally active populations and, second, that a blanket cut-off for 1.55 identifies only about 50% of those defined as under-reporters by EI:EE ratio.7 It ignores those who may have under-reported from high energy requirements, but whose reported intake has not fallen below this conservative cut-off. The conclusions from studies of LERs (reviewed by Macdiarmid and Blundell33) need confirmation by further studies which identify reports of poor validity across the whole range of energy expenditures/requirements. This implies including measurements of activity or energy expenditure as routine in dietary surveys.

At the group level knowledge of activity is needed to assign an appropriate mean PAL value for that population before the presence and/or degree of bias can be determined. At the individual level knowledge of activity or energy expenditure is needed to classify subjects into low, medium or high activity and to assign an appropriate PAL value to each level. It should be noted, however, that this strategy still applies blanket cut-offs across ranges of energy requirements, albeit three narrow ranges rather than one broad range. The direct measurement of energy expenditure is a more desirable aim. Unfortunately, techniques for reporting or measuring activity or energy expenditure are also subject to substantial errors. Detailed discussion of these are outside the scope of this paper. The following paragraphs merely indicate some starting points in the literature.

For large-scale studies, a questionnaire that is simple, easy to administer and easy to analyse is required. Most physical activity questionnaires have been primarily designed to document high intensity exercise or activity sufficient to raise the heart rate. However, much of the variation between subjects comes from differences in time spent sitting, standing and moving around—activities that are difficult to quantify. A questionnaire that elicits the pattern of the general lifestyle, the occupational activity, and the leisure activity is required. One widely used simple questionnaire providing three activity ‘scores' is the Baecke index34 although its usefulness for the particular purpose of evaluating energy intake has yet to be determined. Activity diaries provide more detailed information, but are exceedingly demanding of subject co-operation. They are therefore unsuitable for large studies or those in which compliance may be low. One comparison of a simple questionnaire and a detailed activity diary has suggested that the former might be adequate for large-scale studies.35 On the other hand, a validation of a physical activity recall against DLW energy expenditure and by heart rate monitoring found good agreement at the group but not at the individual level.36

In small studies, simple questionnaires may lack sufficient precision for classification of subjects. It is more desirable to obtain a measure of energy expenditure. Methods for measuring energy expenditure include factorial estimates derived from physical activity diaries, or the proxy measurements of movement by accelerometer or heart rate monitoring. A triaxial accelerometer has been developed that is claimed to account for 65% of the variation in PAL37,38 and may be suitable, in combination with measurement of BMR, to validate energy intake. Heart rate monitoring with individual calibration has also been used to measure energy expenditure in a community-based medium-sized epidemiological study.39 The latter workers are also using heart rate monitoring to validate a physical activity questionnaire which might eventually prove useful as a routine tool.

Conclusions and practical recommendations

  1. 1

    Reported nutrient intake data should not be accepted without critical examination for mis-reporting and consideration of the possible influences of mis-reporting on the conclusions drawn.

  2. 2

    Dietary studies should include an internal validation procedure, whether by measurement of energy expenditure or urinary nitrogen excretion. Validation of energy intake is more appropriate when energy or macronutrients are the parameters of interest. Urinary nitrogen excretion is to be preferred when protein or micronutrients that are highly correlated with protein intake are the parameters of interest.

  3. 3

    Since obesity and dieting have been identified as having the most robust association with mis-reporting, weight and height should be recorded and a short questionnaire to elicit information on weight consciousness and dieting behaviour also included.

  4. 4

    At the group level the Goldberg cut-off can be used to determine the probable degree of overall bias to reported energy intake in a study. Comparison should be made with a PAL value appropriate to the study population based on information about physical activity or lifestyle.

  5. 5

    At the individual level, the use of a single cut-off for EI:BMR applied to all subjects is inappropriate as it fails to identify under-reporters among those with high energy requirements. A short questionnaire on home, leisure and occupational activity such that subjects may be assigned to low, medium or high PAL for calculating the Goldberg cut-off should be included as routine. A short questionnaire is probably adequate and logistically acceptable for large-scale studies.

  6. 6

    In small studies (n<100) it is desirable to obtain a measure of energy expenditure. Detailed activity diaries for factorial calculation of energy expenditure, or possibly the use of triaxial accelerometers or heart rate monitoring are possibilities. If these measurements are obtained, energy intake may be compared directly with energy expenditure and the Goldberg cut-off is irrelevant.

References

  1. 1

    Braam L, Ocke MC, Buenode Mesquita HB, Seidell JC . Determinants of obesity related underreporting of energy intake Am J Epidemiol 1998 147: 1081–1086.

  2. 2

    Briefel RR, Sempos CT, McDowell MA, Chien S, Alaimo K . Dietary methods research in the third National Health and Nutrition Examination Survey: under-reporting of energy intake Am J Clin Nutr 1997 65 (Suppl): 1203S–1209S.

  3. 3

    Ballard-Barbash R, Graubard I, Krebs-Smith SM, Schatzkin A, Thompson FE . Contribution of dieting to the inverse association between energy intake and body mass index Eur J Clin Nutr 1996 50: 98–106.

  4. 4

    Heitmann BL . The influence of fatness, weight change, slimming history and other lifestyle variables on diet reporting in adult Danish men and women aged 35–65 years Int J Obes Relat Metab Disord 1993 17: 329–336.

  5. 5

    Lafay L, Basdevant A, Charles M-A, Vray M, Balkau B, Borys J-M, Eschwege E, Romon M . Determinants and nature of dietary underreporting in a free-living population: the Fleurbaix Laventie Ville-Santé (FLVS) Study Int J Obes Relat Metab Disord 1997 21: 567–573.

  6. 6

    Black AE . Small eaters or under-reporters? In: Ailhaud G, Guy-Grand B. (eds) Progress in obesity research 8th International Congress on Obesity, John Libbey Paris, 1999 pp 223–228.

  7. 7

    Black AE . The sensitivity and specificity of the Goldberg cut-off for EI:BMR for identifying dietary reports of poor validity Eur J Clin Nutr 2000 54: 395–404.

  8. 8

    Haraldsdóttir J, Becker W, Hulshof K, Johansson L, Hermann-Kunz E . Assessment of vitamin and mineral adequacy may be biased by implausibly low energy intakes. Results from five European countries Eur J Clin Nutr 1998 52: S82.

  9. 9

    Macdiarmid JI, Vail A, Cade JE, Blundell JE . The sugar-fat relationship revisited: differences in consumption between men and women of varying BMI Int J Obes Relat Metab Disord 1998 22: 1053–1061.

  10. 10

    Stallone DD, Brunner EJ, Bingham SA, Marmot MG . Dietary assessment in Whitehall II: the influence of reporting bias on apparent socioeconomic variation in nutrient intakes Eur J Clin Nutr 1997 51: 815–825.

  11. 11

    Goldberg GR, Black AE, Jebb SA, Cole TJ, Murgatroyd PR, Coward WA, Prentice AM . Critical evaluation of energy intake data using fundamental principles of energy physiology. 1. Derivation of cut-off values to identify under-recording Eur J Clin Nutr 1991 45: 569–581.

  12. 12

    Rosenbaum M, Ravussin E, Matthews, Gilker C, Ferraro R, Heymsfield SB, Hirsch J, Leibel RL . A comparative study of different means of asessing long-term energy expenditure in humans Am J Physiol 1996 270: R496–R504.

  13. 13

    Prentice AM, Leavesley K, Murgatroyd PR, Coward WA, Schorah CJ, Bladon PT, Whitehead RG . Is severe wasting in elderly mental patients caused by an excessive energy requirement? Age Ageing 1989 18: 158–167.

  14. 14

    Diaz E, Prentice AM, Goldberg GR, Murgatroyd PR, Coward WA . Metabolic response to experimental overfeeding in lean and overweight healthy volunteers Am J Clin Nutr 1992 56: 641–655.

  15. 15

    Black AE, Prentice AM, Goldberg GR, Jebb SA, Bingham SA, Livingstone MBE, Coward WA . Measurements of total energy expenditure provide insights into the validity of dietary measurements of energy intake J Am Diet Assoc 1993 93: 572–579.

  16. 16

    FAO/WHO/UNU . Energy and protein requirements Report of a Joint FAO/WHO/UNU consultation. Technical Report Series 724. World Health Organization: Geneva 1985.

  17. 17

    Black AE, Cole TJ . Within- and between-subject variation in energy expenditure measured by doubly labelled water: implications for assessing the validity of reported dietary energy intake Eur J Clin Nutr 2000 54: 386–394.

  18. 18

    Black AE, Cole TJ. Goldberg GR, Jebb SA, Livingstone MBE, Prentice AM . Critical evaluation of energy intake data using fundamental principles of energy physiology: 2. Evaluating the results of dietary surveys Eur J Clin Nutr 1991 45: 583–599.

  19. 19

    Black AE, Coward WA, Cole TJ, Prentice AM . Human energy expenditure in affluent societies: analysis of 574 doubly-labelled water measurements Eur J Clin Nutr 1996 50: 72–92.

  20. 20

    Department of Health . Dietary reference values for food energy and nutrients for the United Kingdom Report on Health and Social Subjects 41. HMSO: London 1991.

  21. 21

    Edholm OG . Energy expenditure and calorie intake in young men Proc Nutr Soc 1961 20: 71–76.

  22. 22

    Black AE . Dietary energy intake measurements: validations against energy expenditure University of Ulster at Coleraine, DPhil Thesis 1999.

  23. 23

    Black AE, Ravenscroft C, Sims AJ . The NACNE Report: are the dietary goals realistic? Comparisons with the dietary patterns of dietitians Hum Nutr: Appl Nutr 1984 38A: 165–179.

  24. 24

    Black AE . The use of recommended daily allowances to assess dietary adequacy Proc Nutr Soc 1986 45: 369–381.

  25. 25

    Bingham S . The dietary assessment of individuals: methods, accuracy, new techniques and recommendations Nutr Abstr Rev 1987 57: 705–742.

  26. 26

    Nelson M, Black AE, Morris JA, Cole TJ . Between-and within-subject variation in nutrient intake from infancy to old age: estimating the number of days required to rank dietary intake with required precision Am J Clin Nutr 1989 50: 155–167.

  27. 27

    Prentice AM, Goldberg GR, Davies HL, Murgatroyd PR, Scott W . Energy-sparing adaptations in human pregnancy assessed by whole-body calorimetry Br J Nutr 1989 62: 5–22.

  28. 28

    Elia M . Energy expenditure in the whole body. In: Kinney JM, Tucker HN (eds) Energy metabolism. Tissue determinants and cellular corollaries Raven Press: New York 1992, pp 61–79.

  29. 29

    Schofield WN . Predicting basal metabolic rate, new standards and review of previous work Hum Nutr: Clin Nutr 1985 39C: 5–41.

  30. 30

    Schofield W, Schofield C, James WPT . Basal metabolic rate—review and prediction, together with an annotated bibliography of source material Hum Nutr: Clin Nutr 1985 39C (Suppl 1): 1–96.

  31. 31

    Pryer J, Vrijheid M, Nichols R, Kiggins M, Elliott P . Who are the ‘low energy reporters’ in the dietary and nutritional survey of British adults? Int J Epidemiol 1997 26: 146–154.

  32. 32

    Black AE, Bingham SA, Johansson G, Coward WA . Validation of dietary intakes of protein and energy against 24 hour urinary N and DLW energy expenditure in middle aged. women and retired men: comparison with validation against presumed energy requirements Eur J Clin Nutr 1997 51: 405–413.

  33. 33

    Maediarmid J, Blundell J . Assessing dietary intake—who, what and why of under reporting Nutr Res Rev 1998 11: 231–253.

  34. 34

    Baecke JAH, Burema J, Fritters JER . A short questionnaire for the measurement of habitual physical activity in epidemiological studies Am J Clin Nutr 1982 36: 936–942.

  35. 35

    Almendingen K, Trygg K, Pedersen JI . An assessment of the use of simple methods to predict individual energy intakes for intervention studies Eur J Clin Nutr 1998 52: 54–59.

  36. 36

    Racette SB, Schoeller DA, Kushner RF . Comparison of heart rate and physical activity recall with doubly labeled water in obese women Med Sci Sports Exercise 1995 27: 126–133.

  37. 37

    Bouten CVC, Verboeket-van de Venne WPHG, Westerterp KR, Verduin M, Janssen JD . Daily physical activity assessment: comparison between movement registration and doubly labeled water J Appl Physiol 1996 81: 1019–1026.

  38. 38

    Westerterp KR . Physical activity assessment: an objective method to quantify individual energy requirement Eur J Clin Nutr 1998 52: S19.

  39. 39

    Wareham NJ, Hennings SJ, Prentice AM, Day NE . Feasibility of heart-rate monitoring to estimate total level and pattern of energy expenditure in a population based epidemiological study: the Ely young cohort feasibility study 1994-5 Br J Nutr 1997 78: 889–900.

  40. 40

    Wait B, Roberts LJ . Studies in the food requirements of adolescent girls. I. Daily variation in the energy intake of the individual J Am Diet Assoc 1932 8: 323–331.

  41. 41

    Widdowson EM . Study of individual children's diets Medical Research Council Special Report Series 257 HMSO: London 1947.

  42. 42

    McCance RA, Widdowson EM, Verdon-Roe CM . A study of English diets by the individual method. Part III. Pregnant women J Hygiene (Cam) 1938 38: 596–622.

  43. 43

    Edholm OG, Fletcher JG, Widdowson EM, McCance RA . The energy expenditure and food intake of individual men Br J Nutr 1955 9: 286–300.

  44. 44

    Hankin JH, Reynolds WE, Margen S . A short dietary method for epidemiologic studies. II. Variability of measured nutrient intake Am J Clin Nutr 1967 20: 935–945.

  45. 45

    Easty DL . The relationship of diet to serum cholesterol levels in young men in Antarctica Br J Nutr 1970 24: 307–312.

  46. 46

    Balogh M, Kalm HA, Medalie JH . Random repeat 24-hour dietary recalls Am J Clin Nutr 1971 24: 304–310.

  47. 47

    Beaton GH, Milner J, Corey P, McGuire V, Cousins M, Stewart E, de Ramos M, Hewitt D, Grambsch PV, Kassim N, Little JA . Sources of variance in 24-hour dietary recall data: implications for nutrition study design and interpretation Am J Clin Nutr 1979 32: 2546–2559.

  48. 48

    Todd KS, Hudes M, Calloway DH . Food intake measurements: problems and approaches Am J Clin Nutr 1983 37: 139–146.

  49. 49

    White EC, McNamara DJ, Ahrens EH Jr . Validation of a dietary records system for the estimation of daily cholesterol intake in individual outpatients Am J Clin Nutr 1981 34: 199–203.

  50. 50

    Black AE . Pitfalls in dietary assessment. In: Howard AN, McLean Baird I (eds) Recent advances in clinical nutrition :l. John Libbey: London 1981, pp 11–18.

  51. 51

    Bingham SA, McNeil NI, Cummings JH . The diet of individuals: a study of a randomly-chosen cross section of British adults in a Cambridgeshire village Br J Nutr 1981 45: 23–35.

  52. 52

    McGee D, Rhoads G, Hankin J, Yano K, Tillotson J . Within-person variability of nutrient intake in a group of Hawaiian men of Japanese ancestry Am J Clin Nutr 1982 36: 657–663.

  53. 53

    Baghurst Kl, Baghurst PA . The measurement of usual dietary intake in individuals and groups Trans Menzies Found 1981 3: 139–160.

  54. 54

    Stuff JE, Garza C, O'Brian Smith E, Nichols BL, Montandon CM . A comparison of dietary methods in nutritional studies Am J Clin Nutr 1983 37: 300–306.

  55. 55

    Dallosso HM, Murgatroyd PR, James WPT . Feeding frequency and energy balance in adult males Hum Nutr: Clin Nutr 1982 36C: 25–39.

  56. 56

    Soares MJ, Shetty PS . Intra-individual variation in resting metabolic rates of human subjects Hum Nutr: Clin Nutr 1986 40C: 365–369.

  57. 57

    Garby L, Lammert O, Nielsen E . Within-subjects between-weeks variation in 24 hour energy expenditure for fixed physical activity Hum Nutr: Clin Nutr 1984 38C: 391–394.

  58. 58

    Garby L, Lammert O . Within-subjects between-days and-weeks variation in 24-hour energy expenditure at rest Hum Nutr: Clin Nutr 1984 38C: 395–397.

  59. 59

    Soares MJ, Shetty PS . Long term stability of metabolic rates in young adult males Hum Nutr: Clin Nutr 1987 41C: 287–290.

  60. 60

    Soares MJ, Piers LS, Kraai L, Shettey PS . Day-to-day variation in basal metabolic rates and energy intakes of human subjects Eur J Clin Nutr 1989 43: 465–472.

  61. 61

    Henry CJK, Hayter J, Rees DG . The constancy of basal metabolic rate in free-living male subjects Eur J Clin Nutr 1989 43: 727–731.

  62. 62

    Weststrate JA . Resting metabolic rate and diet-induced thermogenesis: a methodological reappraisal Am J Clin Nutr 1993 58: 592–601.

  63. 63

    Figueroa-Colon R, Franklin FA, Goran MI, Lee JY, Weinsier RL . Reproducibility of measurement of resting energy expenditure in prepubertal girls Am J Clin Nutr 1996 64: 533–536.

  64. 64

    Goldberg GR, Prentice AM, Coward WA, Davies HL, Murgatroyd PR, Wensing C, Black AE, Ashford J, Sawyer M . Longitudinal assessment of energy expenditure in pregnancy by the doubly labeled water method Am J Clin Nutr 1993 57: 494–505.

  65. 65

    Goldberg GR, Prentice AM, Coward WA, Davies HL, Murgatroyd PR, Sawyer M, Ashford J, Black AB . Longitudinal assessment of the components of energy balance in well-nourished lactating women Am J Clin Nutr 1991 54: 788–798.

  66. 66

    Schoeller DA, Hnilicka JM . Reliability of the doubly labeled water method for the measurement of total daily energy expenditure in free-living subjects J Nutr 1996 126: 348S–354S.

  67. 67

    Goran MI, Beer WH, Wolfe RR, Poehlman ET, Young VR . Variation in total energy expenditure in young healthy free-living men Metabolism 1993 42: 487–496.

  68. 68

    Haggarty P, McNeill G, Abu Manneh MK, Davidson L, Milne E, Duncan G, Ashton J . The influence of exercise on the energy requirements of adult males in the UK Br J Nutr l994 72: 799–813.

  69. 69

    Bandini LG, Schoeller DA, Edwards J, Young VR, Oh SH, Dietz WH . Energy expenditure during carbohydrate overfeeding in obese and nonobese adolescents Am J Physiol 1989 256: E357–E367.

  70. 70

    Goran MI, Poehlman ET, Danforth E . Experimental reliability of the doubly labeled water technique Am J Physiol 1994 266: E510–E515.

  71. 71

    Goran MI, Poehlman ET . Endurance training does not enhance total energy expenditure in healthy elderly persons Am J Physiol 1992 263: E950–E957.

  72. 72

    Blaak EE, Westerterp KR, Bar-Or O, Wouters JMW, Saris WHM . Total energy expenditure and spontaneous activity in relation to training in obese boys Am J Clin Nutr 1992 55: 777–782.

  73. 73

    Westerterp KR, Meijer GAL, Janssen EME, Saris WHM, ten Hoor F . Long-term effect of physical activity on energy balance and body composition Br J Nutr 1992 68: 21–30.

  74. 74

    Bingham SA, Goldberg GG, Coward WA, Prentice AM, Cummings JH . The effect of exercise and improved physical fitness on basal metabolic rate Br J Nutr 1989 61: 155–173.

Download references

Acknowledgements

I acknowledge with thanks and appreciation the long collaboration and contributions to my own work and thinking of many colleagues but particularly Drs Andrew Prentice, Andy Coward, Susan Jebb, Gail Goldberg, Peter Murgatroyd and Tim Cole at the Dunn Nutrition Centre and Dr Barbara Livingstone at the University of Ulster at Coleraine.

Author information

Correspondence to AE Black.

Rights and permissions

Reprints and Permissions

About this article

Keywords

  • diet record
  • energy intake
  • validity

Further reading